Have a look at this.
It's not so good, actually. Can you believe that it took the artist TEN years of constant drawing practice until she was able to draw this?
I believe it, because I'm the one who drew it. I was eleven years old, and had been drawing all my life because it was fun. Because it was an improvement compared to what I drew at age seven, I used to believe that improvement would magically come by itself if one just kept drawing from their imagination, and doing nothing else. New forms and understandings of anatomy would coincidentally happen to come, little by little. Hadn't it been like that for all those years? From my drawings I could see that I didn't know how to draw an arm correctly, but I believed that if I would draw the arm over and over again from my imagination, the arm would suddenly look better at some point because I had invented a new way to draw it.
Needless to say that improvement is painfully slow when you draw from your imagination exclusively. I met an artist who complained that she didn't improve despite practicing for five years every day; and it turned out that she had drawn from imagination only.
The artist can "know" only so much in their imagination. In order to be able to draw, they have to look at things. You need to draw that arm by looking at an actual figure in front of you. Or a photo. Or a good painting.
I started "copying" when I was thirteen. I made tons and tons of drawings after old master paintings. I attended two figure drawing classes. I copied from my favorite mangaka. I copied tons and tons of photographs in fashion magazines. And suddenly, the improvement didn't come every five years or so, but every year.
Need I say more? Of course, constant life drawing and copying will add tons of new words to your artistic vocabulary, but the human imagination will always stay limited. Take Andrew Loomis and Burne Hogarth, for example. They have drawn so much from life, studied human anatomy for so long, and they are able to draw anatomically perfect figures from their mind. Yet those figures look generic. They are built from the same vocabulary; one instantly thinks "Loomis" or "Hogarth" when they see them. That's why Loomis still used reference for his advertisement paintings and advised the use of it in his books.
Harold Speed in his 1917 book "Drawing Techniques and Materials" expresses the phenomenon like this:
"Try and draw some cumulus clouds from imagination, several groups of them across a sky, and you will find how often again you have repeated unconsciously the same forms. How tired one gets of the pet cloud or tree of a painter who does not often consult nature in his pictures. Nature is the great storehouse of variety; even a piece of coal will suggest more interesting rock-forms than you can invent." (p. 186)
That's the true reason to use reference, I think. Reference isn't a crutch, it's your source for variety. If you like to paint in a realistic style, you can give your paintings a touch of the same variety and uniqueness that nature has, if you reference from nature herself and/or reference photographs.